[Previous] [Next] [Index]
[Thread]
Re: Netscape's purported RNG
> While it might be difficult to do this automatically, it is
> possible for a code review to pick up on things like this. Ross
> Anderson has written some excellent papers (Robustness Principles for
> Public Key Protocols (Crypto '95), Why Cryptosystems Fail (CACM Nov
> 94) and others).
>
> I think there would be a value to a 'Good Security coding'
> seal of approval, implemented by people reviewing code for a price.
Code reviews can pick up some stuff like this, especially if it's an
architectural failing as was the case here. But consider the related
SunOS bug in fsirand(), where a *= was an = instead. It's awfully
hard to see something like that, especially if you're expecting to
see the other. Code reviews are *not* a panacea, as anyone who has
participated in them can tell you.
--Steve Bellovin